Skip to main content

Full text of "Mean field spin glasses treated with PDE techniques"

See other formats





Mean field spin glasses treated with PDE techniques 

Adriano Barraf Gino Del FerraroJ Daniele Tantari ^ 
January 2012 




CIh' Abstract 

Following an original idea of F. Guerra, in this notes we analyze the Sherrington-Kirkpatrick model from 
^y^ I different perspectives, all sharing the underlying approach which consists in linking the resolution of the sta- 

tistical mechanics of the model (e.g. solving for the free energy) to well-known partial differential equation 
(PDE) problems (in suitable spaces). The plan is then to solve the related PDE using techniques involved 
in their native field and lastly bringing back the solution in the proper statistical mechanics framework. 
Within this strand, after a streamlined test-case on the Curie- Weiss model to highlight the methods more 
C/3 ' than the physics behind, we solve the SK both at the replica symmetric and at the 1-RSB level, obtaining 

the correct expression for the free energy via an analogy to a Fourier equation and for the self-consistencies 
with an analogy to a Burger equation, whose shock wave develops exactly at critical noise level (triggering 
j^ ' the phase transition). 

Our approach, beyond acting as a new alternative method (with respect to the standard routes) for tackling 

the complexity of spin glasses, links symmetries in PDE theory with constraints in statistical mechanics 

'■^ ^ and, as a novel result from the theoretical physics perspective, we obtain a new class of polynomial identities 

5IJ . (namely of Aizenman-Contucci type, but merged within the Guerra's broken replica measures), whose in- 

O ' terest lies in understanding, via the recent Panchenko breakthroughs, how to force the overlap organization 

, ^ ^ , to the ultrametric tree predicted by Parisi. 

> ! 1 Introduction 


CN I More than thirty years elapsed since Parisi gave his solution for the mean- field spin glass [3T], namely the 

^^ ■ Sherrington-Kirkpatrick model (SK). During this period, with continuous joined effort from theoretical, numer- 

. I ical and mathematical physics, the community of researchers involved in the field transformed the SK model into 

'"nT ■ the harmonic oscillator for complex systems and features of glassy phenomenology as multiple time sectors [33] , 

^^ I FTD-violations [121 , chaos in temperature [33] , weak/strong ergodicity breaking [HI HO] and last but not least 

ultrametricity |31j are nowadays considerable building blocks of the scaffold of complexity, whose applications 
are spreading over disparate disciplines, ranging from economy |10| and biology [44] to computer science |32| 
and artificial inteUigence |18j . 

However, from a rigorous perspective (i.e. avoiding the zero replica limit within the replica trick framework 
$H ■ [22|). only recently, via a series of impressive results achieved by Guerra |27j and Talagrand |47| and/or within 

the A.S.S. scheme [2j[3l|4T], a crystal-clear formal understanding of the Parisi expression for the free energy has 
been achieved, and even more recently [341 1351 [551 1371 138) . Panchenko has been able to link Parisi ultrametricity 
with polynomial constraints as Ghirlanda- Guerra |25j and Aizenman-Contucci [T] identities, thus conferring a 
central importance to the latter. 

Hence, in the actual state of the art, from one side (and mainly for applications) there is a continuous need 
for new mathematical techniques able to tackle a glassy problem from multiple perspectives and, from another 
(more conceptual) side, a particular focus on the identities (as those we have understood that play a crucial 
role in the ultrametric organization of the states in the low-noise limit of the SK model) raised. 

*Dipartimento di Fisica, Sapienza Universita di Roma, P.Ie Aldo Moro 5, Rome, Italy. 

tDepartment of Computational Biology, KTH Royal Institute of Technology, SE-100 44, Stockholm, Sweden. 

•I-Dipartimento di Matematica, Sapienza Universita di Roma, P.le Aldo Moro 5, Rome, Italy. 

Indeed, despite Panchenko breakthroughs in exploiting their deep hnks with ultrametricity are very recent, 
polynomial identities have a long history in spin glass theory and, since the beginning, the hierarchical flavor 
they contain were manifest (as the title of [23] does not hide): Their early developments are by Ghirlanda 
and Guerra (GG) [5S] and by Aizenman and Contucci (AC) [TJ [22 at the end of the Nineties. Following their 
seminal approaches, the former based on checking the stability of the thermodynamical states by adding all 
possible p-spin terms |201 \n\ to the SK Haniiltonian and then sending their strength to zero (self-averaging 
of the quenched internal energy), the latter obtained trough a property of robustness of the quenched Gibbs 
measure with respect to small stochastic perturbation [TJ [T^l [13] (stochastic stability), identities for the SK 
model have by now been obtained with a number of different techniques, e.g. via smooth cavity field expansion 
[1], linear response stability [IH], random overlap structures [H] or even as Noether invariants |24| . 
Beyond the full mean-field panorama, it has been possible to show to validity of these polynomial identities even 
in short-range or finite-dimensional models |14l[T51[^l46j . however, already restricting in considering the SK- 
model only, the whole identity repertoire is not yet complete and novel techniques to obtain other restrictions, 
with the aim of finding a set of constraints on the overlap probability distribution strong enough to enforce the 
replica symmetry breaking scheme, are still a central focus in spin glass theory. 

As a consequence, by merging the interest in always welcome alternative mathematical approaches to the glassy 
phenomenology with the renewed importance of obtaining novel polynomial identities as they are seen nowadays, 
we decided to continue the investigation of a way initially paved by F. Guerra in solving the thermodynamics 
of the SK-model via well-known PDF techniques, as he first exploited this strand via a bridge connecting the 
SK free energy with the Hamilton- Jacobi theory |28j and then it has been continued in various extensions as 
for instance in [^ [5] [7] [15] . 

The plan for the present paper is as follows: For the sake of clearness, in the next section (Sec. Two), we 
outline the techniques we develop by using the Curie- Weiss model as a toy example. At first, we show that the 
magnetization satisfies a Burger equation whose shock-wave develops exactly at the noise level that triggers the 
phase transition in the SM language, then we introduce its Cole-Hopf transform that maps the evolution of the 
magnetization versus the noise level into a diffusive problem described by a Fourier PDF, which is finally solved 
in a standard way via Green propagator and the Convolution Theorem in the impulse space. As a result of 
this procedure self-averaging of the order parameter is obtained as a bypass product. Then, in Section Three, 
we introduce the Sherrington-Kirkpatrick model that we solve immediately at the replica symmetric (RS) level 
with the same techniques, hence we obtain a Burger equation for the overlap and then we find the shock wave 
that spontaneously develops at the phase transition where annealing breaks down. Via Cole-Hopf transform 
we map the latter into a diffusive problem, that we solve obtaining, beyond the replica symmetric expression 
for the free energy, self-averaging of the order parameter (which is in obvious agreement with the request of a 
RS-solution) and linear constraints of the Aizenman- Contucci type. 

One step forward, we merge this approach with the classical Guerra's broken replica construction [57] to go 
beyond the RS scheme and we show that there is a one-to-one correspondence among the steps of replica sym- 
metry breaking in SM and the spatial dimensions in the equivalent diffusive problem (exactly as happens in the 
Hamilton- Jacobi framework [3) that we then solve in all details at the first step of RSB for highlighting the 
power of the method. 

Remarkably, as a bypass product of this approach, we obtain new identities which constraint overlap fluctua- 
tions on different Guerra measures (roughly speaking constraints between different valleys belonging to various 
steps of RSB). We remark that this kind of identities is completely new and carry interesting physics inside, 
whose discussion, coupled to general outlooks, constitutes the last section, that closes the present paper. 

2 Testing the machinery: The Curie- Weiss model 

Once introduced N Ising spins <Ti = ±1, with i = 1, ..., A^, the Hamiltonian of the Curie- Weiss model can be 
written as 

HN{cr) = -^ 51 '^''^1 ^ - Y"i^' (1) 

where in the last passage (which becomes exact in the thermodynamic hmit N — > oo) we introduced the 

magnetization mj\j = N~^ ^,- ai, namely the order parameter of the theory. 

We introduce also the Boltzmann averages as, using the magnetization as a trial function 

{mN) = 


where the denominator is called "partition function" Z]S[{I3) and of course limjv->oo("^Af) = i^- 

In order to investigate the thermodynamics of the model, we are interested in the mathematical pressure a{(3) 

or equivalently in the free energy /(/3) defined as 

a(/3) = -/3/(/3) = hm ^InZ^iP), (2) 

and we want to achieve an explicit expression for this quantity without following the standard routes of statistical 


The approach we want to use is to "enlarge" the space of the parameters (hence /?), and investigate which 

PDE are obeyed by the model in such a space, so to import the technology for their resolution from classical 

mechanics. In order to exploit our idea, let us introduce the following action S{t, x) as 

a y ij * / 

where the variables t, x can be thought of as fictitious time and space, and such that limAr_j.oo S^it^ x) = S{t, x) 

and of course 5(t, x) — fif{P) = — a(/3) whenever evaluated at i = /? and x = 0. 

In order to highlight our approach we need to work out the derivatives of S{t, x) which read as 

dSN{t,x) 1 


-W- "2^"^^^' 


1 d'^SN{t,x) _ 1 

(toat), (5) 

2N dx^ = -,i(m%)^M')- (6) 

Following the Guerra prescription [551 13 ^^i^d noticing the form of the derivatives (4 — 6), it is possible to 
build an Hamilton- Jacobi equation for SN{t,x) as 

dtSN{t,x) + i {d^SN{t,x)f + V{t,x) = 0, (7) 

where we remark that V{t,x) = —{l/2N)dl^SN{t,x) = (l/2)((m|^,) — {niN)^)'- In the spin-glass counterpart 
this will no longer be the case as polynomial constraints will be present (that here simply reduce to e.g. 
(to'*) = (m^)^, hence are all already accounted by the self-averaging of the magnetization). 
Deriving eq.([7]) w.r.t. x, and cahing UN{t,x) = OxS^it^x) = —{rriN) we get the following Burger equation [35] 
for the velocity (i.e. the magnetization in the statistical mechanics framework apart the minus sign) 

dtUN{t, x) + UNit, x)dxu{t, x) - -rj^dl^UMit, x) = 0, (8) 

and let us point out that such equation becomes naturally inviscid in the thermodynamic limit as 5'(i, x) admits 
the thermodynamic limit thanks to the Guerra- Toninelli scheme |30) . 
If we now perform the following Cole-Hopf transform 

*Af(i, x) = cxp -A^ / dxuAr(t, a;) , (9) 

it is immediate to check that ^N{t,x) satisfies the fohowing diffusion equation 

d'^Nit,x) 1 d-^N{t,x) 

dt 2N dx^ 

= 0, (10) 

which we now soive, in the Fourier space, trough Green propagator and the Convolution Theorem. In the 
Fourier space we deal with ^N{t, k) defined as 

i!N{t,k)^ I dke-'^''-^N{t,x), (11) 

and equation pop in the impulse space reads off as 

5t*Ar(i,fc) + |^*Jv(i,fc)=0, (12) 

whose solution is 

*w(i, k) = *^(0, k) exp i-^t\ , (13) 

and translates in the original space as 

■9Nit,x)^ JdyGtix~y)^oiy), (14) 

where the Green propagator is given by 

Gt{x -y)^ V^exp {~N{x - yf/{2t)) . (15) 

Overall we get 

Sit.x) = -lln^/^ I dy e-^((-y)V(2*)-ln2-lncosh(,)) ^^g^ 

^ ' ■^ N \l 2TTt J " ' ^ ^ 

where we used Sn{0, x) = — log 2 — Incosh(x) and the definition of the Cole-Hopf transform ^. 

As the exponent in equation (|16p is proportional to the volume, for large iV, we can apply now the saddle point 

argument to get 

a{t,x)= sup i - ^ ^' + In 2 + In cosh(2/) \ = - ^ "' + In 2 + In cosh(y ) , (17) 

y y 2t J 2t 

with y maximizer. Note that the previous equation can also be written as 

«(i:a;) + 1^ = sup |-|^ + ln2 + lncosh(j/) + y | = sup |$o(y) + — | , (18) 

hence we obtained the solution also as a Legendre transform of the initial condition^|26|. 
The extremization procedure implies 

X = y — itanh(y) ~ y + u{t,x)t ~ y — {m)t, (19) 

where the second equality holds because the Burger equation becomes inviscid in the thermodynamic limit 
and which, for x = (where statistical mechanics is recovered) implies {/q ~ {m)t so to obtain the well-known 
Curie- Weiss self-consistency (properly evaluated by choosing t = f3) 

(m) =tanh(/3(m)). (20) 

^Strictly speaking $0(3/) = ~J/^/2t -I- ln2 -|- lncosh{j/), hence is the initial condition plus —y^/2t. 

The free energy of the Curie- Weiss model is then 

a{P) = sup I In 2 + In cosh (/3(m))- §(m)H . (21) 

(m) I 2 J 

While it is well known from classical arguments of statistical mechanics that the Curie- Weiss model undergoes 
a phase transition from an ergodic (paramagnetic) phase to a ferromagnetic one at /3 = 1, it is very instructive 
to tackle this phenomenon still within our framework, where such a phase transition is obtained as a shock wave 
for the Burger equation ([5]). 

In order to see this, it is useful to investigate the mass conservation, whose density is depicted by the variable 
p, namely by analyzing the mass of the "fictitious particle" whose motion we study in two different positions, 
a generic x and the starting point y, as 

p{x)dx = p{y)dy. (22) 

By the equation of motion x = y + u(0, y)t ~ y — tanh(j/)i we get 


— = 1 + dyu{Q, y)t = 1-{1- tanh2(y))i, (23) 


thus, for the mass density in a generic point x, we get 

At y = 0, p{x) = p(0)/(l — t) which diverges for t = 1, thus, as t = /3, exactly where the phase transition 
happens in statistical mechanics. 

Note further that in the Hamilton- Jacobi equation ^, the potential V{t, x) is vanishing for A^ -> oo as V{t, x) = 
d^^S{t, x)/2N, and the corresponding Burger equation ([S]) becomes inviscid: As a result, by definition of V{t, x), 
we get self-averaging of the order parameter, namely limN-^ooiiM^) — (Mn)'^) = 0. In the spin-glass counterpart 
this procedure will develop more complex overlap polynomial identities and, while even in that context we will 
have \iinN^aoV{t,x) = and lim^v-i-oo 9^i^5(t, a;)/2A^ = 0, the two results will be in general different, the 
former representing overlap self-averaging hence restricted only to the RS scenario, the latter representing more 
general polynomial identities (for simple models as the CW, these two results of course do coincide -as they 
should- because the model is intrinsically replica symmetricp). 

3 The Replica Symmetric Sherrington-Kirkpatrick model within the 
Fourier framework. 

Once introduced A^ Ising spins ai = ±1, the Hamiltonian of the SK model is given by 

HN{cr; J) = j= Y^ .Aja.aj (25) 

^ r ■^ 

where the quenched disorder in the couplings is given by the N{N — l)/2 independent and identical distributed 
random variable Jij, whose distribution is A/'[0, 1]. 

We are interested in an explicit expression for the (quenched) free energy /(/3) (or the mathematical pressure 
a(/3)) defined as 

a(/3) = -/3/(/3) = lim a^(/3) = - lim ^/nW) = Jim ^Eln Zn{P), (26) 

A*— >oo A*— >-oo A/— foo I\ 

where E averages over the quenched couplings and Z]y(/3) = exp{—l3H]^{a; J)) is the partition function. 
Through Zi^{j3) we define the Boltzmann state w(.) = ^^ . exp(— /?iJAr(cr; J))/(ZAr(/3)), the product state 

^For the sake of completeness, note that the Parisi-hke representation of the CW model is shown in [5]. 

ri(.) = w(.) X ... X uj{.) and the averages (.) = £!!(.). 

Now, mirroring the previous section, we introduce two fictitious variables x,t, which can be though of as space 

and time coordinates, by which we write the Guerra's interpolating function as 

aN{t,x) = — Eln^exp J j;^^Jtj(^t<^3 +Vx^Jl(^t , (27) 

<T y i<j i j 

where the J^^'s are i.i.d. unitary gaussian random variables, and the pressure is recovered whenever evaluating 
aAr(i, x) at i = /3, a; = 0. Further, as a is not directly connected to an Hamilton- Jacobi equation; we need a 
linear transformation in the i, x plan to introduce the Guerra's action Stqit^x) as 

S'iv(i,x) =2ajv(^,a;)-x-i/2. (28) 

By direct calculation, we can see the following relations holding [5H] 

(9tS'jv(i,x) = --(g?2), .29) 

dxSfqit.x) = - (qi2) , 

where we implicitly introduced the overlap, e.g the order parameter of the theory, defined as qyi = 7V^^ ^^ '^I'^fj 
whose A^-dependence has been omitted for the sake of simplicity. By direct construction it is immediate to check 

dSNit,x) 1 fdSN{t,x) 

) =-l{{ql2)-{<}i2r)^-Voit,x). 

dt 2 \ dx 

If we add a vanishing (in the thermodynamic limit) potential, containing the second derivative of SNit,x) 


1 (PSN{ t,x) 

n'-^'oo 2N dx 

li- T^^^^^^^^^V,{t,x)=0 (31) 

and within the replica symmetric scheme, where limjv-i-oo (('Z12) ^ (912)^) = 0, S'jv(^, x) satisfies 

lim {dtS^it, x) + l {d.S^it, x)f - ^dl^S^it, x)) = 0, (32) 

that we can solve easily with the usual Cole-Hopf transform (see the next section) . 

A remark is in order here: As discussed in detail in J3, and as of course the overlap is not self-averaging in the 
true solution of the SK model (21], we force Vo{t,x) to be zero in order to get straightforwardly the replica- 
symmetric solution (which is the goal of the present section), while Vi{t, x) is always zero in the thermodynamic 
limit (and of course reduces to an elementary identity once read in the RS framework). 

Note that, while it is not strictly necessary to solve this problem where Vo(t, x) and Vi{t,x) are pasted in 
the same equation as we could split the standard Hamilton- Jacobi equation for the Guerra action from the 
constraint 2^ — ^2 '^ = 0, however, such a "compact procedure" allows to obtain the RS free energy solving a 
Fourier problem (with all its related know-now) for its Cole-Hopf transform. 

To compute explicitly Vi (t, x) it is convenient to introduce the x-streaming relative to a generic observable F 
which depends on s replicas as [5S] 

d, (F,) = N{F Ij2l-b- sE'?-+i + ""^''^^^ ^+1,^+2)). (33) 

\ ab a / 

hence, remembering that, from (P^ . dxS{x,t) = — ((Z12), we get 

lim ^dl, S^it, x) = lim i{qf,) - 4 (^12923) + 3 {qms^)) = 0. (34) 

The latter is a constraint between overlap's polynomials known as AC identities [T] and in particular it coincides 
with eq.s (47,48) of [55], where this kind of identities appeared for the first time, previa elimination of (912)^ 
to reduce to a single expression. Hence, as a difference with the CW, Vo{t,x) ^ Vi{t,x) but they both vanish 
in the N ^ 00 limit (Vg only in the RS framework) and they carry different physics inside (as we are going to 
deepen later, in the broken replica framework). 
As we did for the CW model, we can solve the Burger-like equation for the action 

dtSN{t,x) + ^{d,SN{t,x))^ - ^dl.SN{t,x) = (35) 

mapping the latter into a Fourier equation via the following Hopf-Cole transform, namely 

*Ar(i, x) = cxp {-NSN{t, x)) (36) 

by which it is straightforward to check that *Ar(t, x) obeys: 

d^{t,x) 1 d^^{t,x) _^ 

dt 2N dx^ ~ ^ ' 

By calling *I'Ar(i, k) the transform of ^Ar(t, x), in the impulse space we get 

dti>N{t,k) + ^i>N[t,k) (38) 

hence, using the label ^o(^) to denote the Cauchy condition, we get the solution in the /c-space as 

*Ar(<,fc) = 'J'o(fc)cxp(-— t). (39) 

We can obtain the solution in the original space by the Convolution Theorem, hence writing 

^{t,x) = JdyGtix~y)M>o{y)^^£-^Jdye-^^^^^o{y) (40) 

where Gt{x — y) is the Green propagator, i.e. the solution of p7p with the initial condition Go{x) = 6{x) and 

*o(y)-cxp[-7V5o(2/)]. (41) 

From the definition of the interpolating function (P7)) and from the definition ([55)) one can directly get the 
expression for So{y), namely 

5o(y) = 21n2 + 2 f dfi{z) lncosh(Vy2) - y, 
and by direct substitution we get 

SM{t,x) ^--log^it,x)^--log^— I dye V^' ^ J (42) 

that in the thermodynamic limit reads 

5(i,x-)-inf|^^-^ + ^o(2/) 

= mil — — — + 21n2 + 2 / d^l{z) In cosh{^z) ~y\ , for TV -^ 00. 

We denote by y{t,x) the location where the infimum in (|43|) is achieved and refer to this function as inverse 
Lagrangian function [26| . As we are going to show, its inverse x{t, y) is the usual Lagrangian function, or rather. 

the location at time t of the fictitious particle initially at y. We observe that the previous maximizing condition 
can also be expressed as: 

s{t,x)-^ = -sMMy) + Y^ (44) 

where (j>Q{y) = —y'^/2t~So{y). Hence, the solution of the Burgers-like equation can be expressed again in terms 
of a Legendre transform of 4>a{y). From the extremization condition we get 

X = y — t / (i/i(z) tanh {\/yz) = y + tu{t, x), (45) 

with y maximizer. The last equality is allowed because in the thermodynamic limit the Burger equation 
becomes inviscid, hence trajectories represent Galilean motion with a velocity u(t, x) given explicitly by the 
previous expression. 

Under the replica symmetric assumption, from equation H29p and the definition of u{t,x) — dxS(t,x), in the 
thermodynamic limit we can finally recover the self-consistent equation for the overlap 

i{t,x) = d^S{t,x) = - {qi2) it,x) = -qit.x) = - / dti{z)t&nh^{^/y{t,x)z). (46) 

To recover statistical mechanics we need evaluating observables at x = (e.g. in equation (1451) ') and the value 
of yo = tq that maximizes the expression (j44p is then 

yQ = tq = t / dfi{z)tanh.^{y^z). (47) 

By considering equation (j43p . with y maximizer, we have 

S{x,t)= ^^~^^ +21n2 + 2 f dfi{z)lncosh{y^z)-y, (48) 

and when evaluating everything at a; = and i = /3^ (thus we use the relation yo = 9/?^) we finally get 

a(«^,„2 + /dM.)l.>c„sh(V?^.)4(l-,-f (49) 

which is the expression of the pressure for the SK model in the RS approximation. 

As in the Curie- Weiss model, we want to investigate the shock waves in the Burgers equation and figure 
out some correspondences with the phase transition in statistical mechanics (which is known to happen at 
/3 = 1). To achieve this result, mirroring the corresponding section for the ferromagnetic case, we consider the 
conservation of the mass for a particle which starts the motion in position y and arrives at position x after a 
time t: 

p{x)dx = p{y)dy. (50) 

Reminding the equation of motion x = y + m(0, j/)i, from mass conservation we get 

''W-'''(S)""^TT7^- '-' 

From equation psj) we get 

dx If 

— = l + tdyu{0,y) = l~t—dfi{z)z=^f^ (52) 

dy y/y J cosh^ {^z) 

If we want to find a shock we can evaluate miuj, 

tanh(yy z) 

, which can be achieved simply 

1 r^,,/..\., tanh(Vy z) 

observing parity and monotony in y G [0,oo) hence obtaining y ~ 0. Then if we expand the tanh for small y 
we obtain 

^2 2 f°° Z^ 1 

2/ dp,{z) ^2/ ^ . +^y / d^i{z) 2, ^ ■. == ~1' (53) 



cosh {y/yz) 3 Jo cosh {^z) 

where the final result is obtained computing the argument inside the brackets for y = y = 0. We can then 
conclude that dyu{Q,y)\y = — 1. If we now substitute this last result in eq ([?T|) we get 

oM = ^^ = ^ (54) 

^^"^^ l + dyuiO,y)\yt 1-t ^''*> 

which diverges at the shock time t = /3^ = 1, i.e. the critical noise level at which the phase transition occurs in 
statistical mechanics. 

4 Beyond the replica symmetric scenario: Multiple diffusion and bro- 
ken replica constraints. 

In this section we want to go beyond the replica-symmetric scenario and investigate features of the broken replica 
phase by merging the PDE approach, in particular the Fourier technique, with the broken replica interpolation 
scheme developed by Guerra in |27) . The result will be a mapping between the SK free energy with some steps 
of RSB (that we are going to analyze in full detail for the 1 — RSB test-case for the sake of simplicity, mirroring 
the work developed in [7] for the Hamilton- Jacobi procedure), and a multi-dimensional diffusion equation, with 
as spatial dimensions as the number of RSB steps plus one, hence two spatial dimensions in order to tackle the 
1 - RSB solution. 
To accomplish this task let us introduce the following interpolating partition function 

CN N \ 

i i / 


where j}, Jf are all i.i.d. Gaussian variables, sharing the same distribution A/'[0, 1] as the original J^ but 

independent from them, whose averages will be denoted with the pertinent subscript for the sake of clearness, 

hence E2, Ei. Moreover we have to think at to G [0,1] as the expected parameter of the 1-RSB overlap's 

distribution, i.e. we assume its shape to be the weighted sum of two delta functions, P{q) = ■m6{q — qi) + (1 — 


Let us introduce recursively the "partially averaged" partition functions as 

Z2 = Zn, Zr^E2{Z'^), 
by which we can express the 1-RSB SK free-energy (or pressure strictly speaking) in the space {t,xi,X2) as 

a(t, xi,X2)~ lim aN(t,xi,X2) — lim -— EEi log Zi(i, xi, 2:2) = hm -— — EEi logE2Z]^(i, a;i,X2). (56) 

N~¥oo Af-i-oo N JV-i-oo Nm 

Trough the interpolating structure defined by the extended partition function (j55p and the partial averages Ej , 
we can introduce also the following weights / and the corresponding extended states uj as 

h ^1, h- ^, (57) 

ioi{.)^M.hi^2{-)l W2(.)=^Jv(.), (58) 

while the standard product state remains fla{-) ~ (^a{-) x •■• x '^a(-) and finally the quenched averages read 

{.)a=E{h...fMa{.)), (59) 

hence (.)i = E[rii(.)], and {.)2 =E[f2n2{.)]. 
If we now define 

SN{t, xi,X2) = 2aNit,xi,X2) - X2- t/2, (60) 

we can see that the following relations hold: 

dtSN{t,Xi,X2) = -{m/2){ql2),-{{l-m)/2){ql2)2, 
dx2SN{t,xi,x2) = -(1 - "l) (912)2, 
dxiSNit,Xi,X2) = -m(qi2)i, 

by which we can see that SN{t,xi,X2) satisfies 

^'^^ + i (5-1^^)' + 2(^1^) id,,,SNf = -i [m [{ql,)^ - (gi2>?) + (1 - m) ((g^^)^ - (912)^)] . (6f) 

In the f-RSB approximation we can neglect the r.h.s. of (|6ip . moreover, if we add a vanishing term containing 
the second derivatives of SN{t, xi,X2) we can write down a Burger equation for S(t, xi,X2) as 

that we can solve by mapping it into a Fourier equation via the usual Cole-Hopf transform 

*Ar(t, a;i, 2:2) = cxp {-NSNit, a;i, 3:2)) , (63) 

by which it is straightforward to check that ^Ar(t, xi, 2:2) obeys 

hmJdt^r,{t,x,,X2) - ^^^92,vl>^(t,xi,x2) - ^]^^;^^9l,^N{t,x,,x,)) = 0, (64) 

i.e. a diffusion equation with two different diffusion coefficients on the two spatial axes xi, X2, namely Di = 
{l/2Nm) and D2 = 1/(2A^(1 - m)). 

Now we want to solve the heat-equation (j64p in the Fourier space, where, calling ^Ar(i, fci, ^2) the transform of 
^Ar(t, xi, X2) we can write 

y^2 ^2 

at*w(t, fci, A:2) + 7r^*w(t, fci, fc2) + 7T7^^-Tl^*iv(i, fci, fc2) = 0, (65) 

IrnN 2(1 — rn)iV 

hence, using the label ^0(^1, ^2) to denote the Chauchy condition, we get 

\2mN "*" 2(l-m)iV 

Finally we can obtain the solution in the original space by the Convolution Theorem, hence writing 

f dki f dk2 - f f k^ k^ \ \ 

'^Nit,xi,X2) = / — == / — ==*o(^, fci,fc2)exp - — ^ + — - — '^—-—]t]cxp{ikiXi+ik2X2){67) 

*7v(t, fci, fc2) = ^o(fci, fc2) cxp ~ —^ + ^.^ ' ,^^ ]t]. (66) 

^y ^2^^uv",.".,."z;^^^^^ ^2m7V 2(l-m)Ar^ 
dyi dy2'i>oit,yi,y2)Gtixi-yi,X2-y2), (68) 

where Gt{xi — yi, X2 — 1/2) is the Green propagator, i.e. the solution of (|64p with the initial condition Gq{xi, X2) 

GAxi,X2) = [^ f^e l^+^a-*^J*e(*.i+^fc...) ^ Ag-A^d-J+^i^-^), (eg) 

^ ' ' J 2n J 27r 27rt ' v ; 

by which the solution of eq. ((64)) is 

5w(t,a;i,X2) = -^logvI/(t,xi,X2) = -^logydyiydy2e-^(^«('^^^^=)+^("^-^^>'+^^^("=-^=)'). (70) 


In the thermodynamic hmit we can use the saddle point method to get 

(1 — m) 
S{t, xi,X2)^ inf {So{yi,y2) + ^{xi-yif -\ — — {x2 - y-i)'^ } , 



and finally, keeping in mind the definition ((60 

a[t,xi,X2)= mf <-So[yi,y2) + -rT[xi-yi) H — (X2 - y2) 

yi,V2 [2 At 4i 

2 1 ifi I i 

2 4 



At i = we have 

If the minimum of the equation (|7ip is achieved in (j/i, 2/2) = (j/i, 2/2), we can write 

Soiyi,y2) = 2a{0,yi,y2) - y2 = 21og2H / rf^(2i)log / d/^(z2) cosh™ (7^1^:1 + Vy2 - 2/1^2) - y2- (73) 



ax I 

-m{qi2)i = m 

xi -yi{t,xi,X2) 




-(1 -m) (912)2 = (1 -m) 

X2 - y2{t,Xi,X2) 

by which we get the physical meaning of the variables j/i , y2 as 

yi(i,xi,X2) = xi + {qi2{xi,X2,t))it, 

y2it,Xi,X2) = X2 + {qi2ixi,X2,t))2t. 



This means that, at i = /3^ and xi = X2 = 0, where we come back to the original SK model, {yi,y2) — 

/3'((9i2)i, (912)2) =/3'(gi, 92). 

Now we want to extremize the expression inside the brackets of (|72p . Taking the derivatives w.r.t. y\ and 
1/2 , by straightforward calculation we get 

xx -'yi{t,Xi,X2) 

X2 ~y2{t-,Xi,X2) 

D ^{zi,y) dfi{z2)cosh"'{B{y,zi,Z2))tauh{&{y,zi,Z2)) 


d^i{zi)D \zi,y) / d^(2;2)cosh'"(e(y,zi,Z2))tanh2(e(y,zi,Z2)), 


where we defined 


d^{z2) cosV^^/^zi + yjy2 -yiZ2) 

v^^i + ^m ~yiZ2, 

and by which we can reconstruct the function a(t, xi, X2) as 


a[t,xi,X2) = ln2 + —{xi-yi{t,xi,X2)) H -— 

2 , ^ ^ "V ^ -.2 X2 - y2 t 

{X2 - 2/2) + ^ + 7 

+ — dfi{zi)\n / (i/i(z2)cosli'" f v^^i + Vv^ ~ y2Z2J ■ 


At i = /3^, xi =0:2 =0 and identifying (^1,^2) = /3^(9i,92) we find the following well known [2Z][Z] self- 
consistent equations 


92 = 


n 2 

D (21,9) / d^i{z2) cosh"(^V9i-2i + /^\/92 - 91^2) tanh(/3V9i2:i + P^/q2 - 912^2) 

d^{zi)D (21,(7) / ii/i(22)cosh™(^yg72:i +/3V92 - 9i22)tanh (/3V9i"-2i + /3\/92 - 912^2) (81) 


and the 1 — RSB free energy of the SK model as 

a(/3) = ln2+^ [mql + (1 - m)ql - 2q2 + l) + ^ f dfi{zi) \n f dfi{z2) cosh'"' (/? (^/|^Zl + V92 - 9122)) • (82) 

Earlier, in writing the equation (j62p . we added a vanishing (in the thermodynamic limit) term concerning 
the second derivatives of S{t, xi,X2)'- 

N(sb)«i^ = >.(,.). "^-0. (84) 

These terms are vanishing because S{t, xi, X2) is a continuous function in its parameters and its thermodynamic 
limit exists due to the Guerra-Toninelli argument |30| or, from a more practical point of view, as we have found 
the correct 1-RSB free energy, they necessarily have to be vanishing in the thermodynamic limit. Analyzing the 
equations (|83l84p allows us to find some overlap constraints that generalize the AC like identities found inside 
the RS scheme (equation ([M)l V Let us start exploiting the meaning of the first equation. Let us stress that 
^2{-) = W2(.) X ... X L02{-) = ^^(Oj while Vti{.) = Cl'i(.) x ... x a;i(.), furthermore, for the sake of clearness, let us 
remember that (.)2 = E[/2ri2(-)] ^^'^ (•)i = IE[rii(.)]. We need further the introduction of the state 

that is not a replicated state because couples different replicas with the E2 average. Finally we have to define 
composite states: without using an unnecessarily general notation, we will define for example 


E[(Oi X r!i)(.)] (85) 

and similarly for any other possible combination of replicated states. It is convenient to introduce the following 
preliminary results: 

/"" 1 

9,72/2 = "^]g^^ aj2Z = mf2^yx2 " xicr,), (87) 

dj2Uj{F) = yjx2 — x\ {uj{Faj) - ui{F)ui{uj)) , 

d.,MF) = ^ J Y. Jk {^{Fau) ~ u{F)u{au)) . (89) 

With these premises, the evaluation of the streaming of (912)2 over X2 can be written as 

d^, (512)2 = E (9,,/2f^(gi2)) + E (/2a,,17(gi2)) =21 + 2, (90) 


/ 777 1 777 " — ill \ 

21 = iV (777(912923)2 + ^ (912934)2 + 2™(1 - H2(9i2934)nixfii j > (91) 

» = iV((TO-4)(9i2923)2 + (9?2)2 + (3-TO)(9l2934)2), (92) 

hence, finally the first constraint reads off as 
Jim —5^^2(912)2 =0 = (9^2)2 + 2(m-2)(9i2923)2 + -(m-3)(m-2)(9i2934)2 + xm(l-m)2(9i2934)oixni- (93) 


In the same way we get the other constraint that follows from equation (|84p by which, overall, we get 

= (9i2>2 + 2(m-2)(gi2g23>2 + :r(m-2)(m-3)((ji2g34>2 + -m(l-TO)(gri2934)oixni ' (94) 

= (gDi -4m(9i2g23>i +3m^ (912^34) 1 +2(m- 1) (gi2gi3 

2' " /N^--.o,,. 2 

o; 1 X 1 

+ 4m(l - m) (913924)^1x121 + (1 - "^)^ (9i3924>Oixsii + 2?7i(m - 1) (912934)^1x01 (95) 

Note that, as expected, when 771 = the first constraint becomes as the AC like identity ([M)l . derived using the 
RS interpolating scheme, but concerning the state {.)^, i.e. averaging only inside the first ensemble of valleys. 
On the contrary, when m = 1, the second constraint becomes the AC like identity ([M|) but with the state 
(.),, i.e. averaging just inside the second ensemble of valleys. When m E (0,1) other terms appear to take 
into account the relation between the two valleys, hence constraining, for the first time, this kind of overlap 

5 Conclusion and outlooks 

In this paper we highlighted how standard PDE techniques, well consolidated in different fields of mathemat- 
ical research (e.g. classical mechanics and hydrodynamics), may work as well once tested on the disordered 
statistical mechanics machinery. We extended previous results ESI [3 [Z] EH on this line, showing how to solve 
trough Fourier theory for the free energy and via a Cole-Hopf transform by using Burger theory for the order 
parameter. Remarkably, as the latter admits shock waves, we found that -within this parallel- those happen 
exactly where phase transitions develop in standard statistical mechanics. 

Beyond showing the main ideas on the Curie- Weiss model for the sake of clearness, we tested our approach 
to the paradigmatic Sherrington-Kirkpatrick model, which has been solved both within a replica symmetric 
scenario and within the Guerra broken replica framework, at the first level of replica symmetry breaking for the 
sake of simplicity, finding full agreement with the classical theory. 

We remark however that, while general iiT-steps of RSB make the calculations only longer, but work straight- 
forwardly, the K ^ 00 limit requires still some effort on which we will deserve future investigations. 
A main point to highlight is that our procedure is a novel scheme to produce (polynomial) overlap constraints, 
which have been recently found to play a crucial role |34l 155] in the ultrametric organization of the states 
predicted by Parisi theory |31) and that we obtain as a constraint on the theory. To get an entirely class of new 
identities, we used a combination of results: as those which could be derived within the standard Gibbs measure 
-in the fully connected topology underlying the SK model- have been already obtained (and we found them 
as well in the simpler RS-framework) we worked out this kind of constraints within the Guerra broken replica 
weights. This procedure resulted in correlations that link overlap fluctuations between different measures, which 
roughly resembles constraints among valleys of different depth (corresponding to different steps of RSB from 
an heuristic perspective). We remark that these constraints have been obtained very naturally, from continuity 
arguments, as the second (spatial) derivative of the action, once divided by the volume, must go to zero in 
the thermodynamic limit: Roughly speaking, this approach resembles self-averaging in older investigations |25j . 
and, forcing the broken-replica scheme to collapse on the simple Gibbs measure (hence imposing m = I), results 
reduce to those well known polynomial identities (already found in [TJ [23 HI)- Interestingly in this way, we 
obtained an entirely new class of identities that constraint overlap fluctuations within different levels of RSB: 
The implication in possible closure of the SK-repertoire, hence in supporting Panchenko results will be subject 
of future investigation as well. 


The authors are indebted with Erik Aurell for several illuminating discussions on the shock wave and the inverse 


Francesco Guerra is warmly acknowledged as usual for his priceless constant guide. 

AB and DT acknowledge the FIRB grant RBFR08EKEV and Sapienza Universita' di Roma for partial financial 

support and GDF acknowledge a grant by Netadis Project. 



[1] M. Aizenman, P. Contucci, On the stability of the quenched state in mean field spin glass models^ J. Stat. 
Phys. 92, 765, (1998). 

[2] M. Aizenman, R. Sims, S. L. Starr, An Extended Variational Principle for the SK Spin-Glass Model, Phys. 
Rev. B 68, 214403, (2003) 

[3] M. Aizenman, R. Sims, S. L. Starr, Mean-Field Spin Glass models from the Gavity-ROSt Perspective, 
Prospects in Math. Phys. 437, J.C. Mourao, J. P. Nunes, R. Picken, and J-C Zambrini (eds.) (2007). 

[4] A. Barra, Irreducible free energy expansion and overlap locking in mean field spin glasses, J. Stat. Phys. 
123, 601, (2006). 

[5] A. Barra, The mean field Ising model trough interpolating techniques, J. Stat. Phys. 132, 787, (2008). 

[6] A. Barra, L. De Sanctis, Overlap fluctuation from Boltzmann random overlap structure, J. Math. Phys. 47, 
103305, (2006). 

[7] A. Barra. A. Di Biasio, F. Guerra. Replica symmetry breaking in mean field spin glasses through the Hamilton 
Jacohi technique, J. Stat. Mech. P09006, (2010). 

[8] A. Barra, G. Genovese, F. Guerra, The replica symmetric approximation of the analogical neural network, 
J. Stat. Phys. 140, 784, (2010). 

[9] J. P. Bouchaud, D.S. Dean, Aging on Parisi tree, J. de Physique I 5, 265, (1995). 






J. P. Bouchaud, M. Potters, Theory of financial risk and derivative pricing: from statistical physics to risk 
management, Cambridge University Press (2000). 

A. Bovier, P. Picco, Mathematical Aspects of Spin Glasses and Neural Networks, Birkheauser Editor, (1998) 
and references therein. 

P. Contucci, Stochastic Stability: A Review and Some Perspectives, J. Stat. Phys. 138, 543, (2010). 

P. Contucci, C. Giardina, Spin Glass Stochastic Stability: A rigorous proof, Annales Herni Poincare, 6, 
915, (2005). 

P. Contucci, C. Giardina, The Ghirlanda- Guerra identities, J. Stat. Phys. 126, 917, (2007). 

P. Contucci, C. Giardina, C. Giberti, Interaction- flip identities in spin glasses, J. Stat. Phys. 135, 1181, 

P. Contucci, C. Giardina, C. Giberti. Stability of the Spin Glass Phase under Perturbations, Europhys. 
Lett. 96, 17003, (2011). 

A.C.C. Coolen, The Mathematical theory of minority games - statistical mechanics of interacting agents, 
Oxford University Press, (2005). 

A.C.C. Coolen, R. Kuhn, P. SoUich, Theory of neural information processing systems, Oxford University 
Press, (2005). 

L.F. Cughandolo, J. Kurchan, L. Pehti. Energy flow, partial equilibration, and effective temperatures in 
systems with slow dynamics, Phys. Rev. E 55, 3898, (1997). 

B. Derrida, Random energy model: An exactly solvable model of disordered systems, Phys. Rev. B 24, 2613, 

B. Derrida, E. Gardner. Solution of the generalized random energy model, J. Phys. C 19, 2253, (1986). 



















V. Dotsenko, One more discussion of the replica trick: the example of the exact solution^ Phil. Mag. 92, 1, 
16, (2012). 

S. Franz, M. Leone, Replica hounds for optimization problems and diluted spin systems, J. Stat. Pliys. Ill, 
535, (2003). 

G. Genovese, A. Barra, A mechanical approach to mean field spin models, J. Math. Phys. 50, 053303, 

S. Ghirlanda, F. Guerra, General properties of overlap distributions in disordered spin systems. Towards 
Parisi ultrametricity, J. Phys. A 31, 9149, (1998). 

S.N. Gurbatov, S.I. Simdyankin, E. AureU, On the decay of Burgers turbulence, J. Fluid Mech. 344, 339, 

F. Guerra, Broken replica symmetry bounds in the mean field spin glass model, Comm. Math. Phys. 233, 
1, (2003). 

F. Guerra, Sum rules for the free energy in the mean field spin glass model. Mathematical Physics in 
Mathematics and Physics, Field Inst. Comm. 30, 161, (2001). 

F. Guerra, About the overlap distribution in mean field spin glass models, Int. Jou. Mod. Phys. B 10, 1675, 

F. Guerra, F. L. Toninelli, The Thermodynamic Limit in Mean Field Spin Glass Models, Comm. Math. 
Phys. 230, 71, (2002). 

M. Mezard, G. Parisi, M.A. Virasoro, Spin glass theory and beyond, World Scientific, Singapore, Lect. 
Notes Phys. 9, (1987). 

M. Mezard, A. Montanari, Information, physics and computation, Oxford University Press, (2009). 

J. van Mourik, A.C.C. Coolen, Cluster derivation of Parisi 's RSB solution for disordered systems, J. Phys. 
A 34, Llll, (2001). 

D. Panchenko, Ghirlanda- Guerra identities and ultrametricity: An elementary proof in the discrete case, 
C. R. Acad. Sci. Paris 349, 813, (2011). 

D. Panchenko, A unified stability property in spin glasses, to appear in CMP, available at! arXiv:1106.3954l 

D. Panchenko, The Parisi ultrametricity conjecture, Ann. of Math. (2), 177, 1, 383, (2013). 

D. Panchenko, A unified stability property in spin glasses, Comm. Math. Phys. 313, 3, 781-790, (2012). 

D. Panchenko, A connection between the Ghirlanda- Guerra identities and ultrametricity, Ann. of Prob. 38, 
1, 327, (2010). 

T. Rizzo, A. Crisanti, Ghaos in temperature in the Sherrington- Kirkpatrick model, Phys. Rev. Lett. 90, 
137201, (2003). 

D. Sherrington and S. Kirkpatrick, Solvable model of a spin glass, Phys. Rev. Lett. 35, 1972, (1975). 

P. SoUich, A. Barra, Notes on the polynomial identities in random overlap structures, J. Stat. Phys. 147, 
351, (2012). 

Z. She, E. AureU, U. Frisch, The inviscid Burgers equation with initial data of Brownian type, Comm. 
Math. Phys. 148, 623, (1992). 

S. Starr, Thermodynaic limit for the Mallows model on S'„,, arXiv:0904.0696, (2009). 


[44] D.L. Stein (Ed.), Spin glasses and biology, Singapore, World Scientific (1992). 

[45] S.A. KaufFman, The origin of order, Oxford University Press (1993). 

[46] M. Talagrand, Spin glasses: a challenge for mathematicians. Cavity and Mean field models. Springer Verlag 

[47] M. Talagrand, The Parisi Formula, Annals of Mathematics 163, 221, (2006).